power aware manhattan routing on chip multiprocessors
play

Power-aware Manhattan routing on chip multiprocessors Anne Benoit 1 - PowerPoint PPT Presentation

Framework Theoretical results Heuristics Simulations Power-aware Manhattan routing on chip multiprocessors Anne Benoit 1 , Rami Melhem 2 , Paul Renaud-Goud 1 and Yves Robert 1 , 3 1 . Ecole Normale Sup erieure de Lyon, France, {


  1. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p ) . Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 11 / 34

  2. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p ) . K : sum of all communications K (1) k : the sum of the γ i that cross D (1) k In this case, K (1) = K for each k k Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 11 / 34

  3. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p ) . K : sum of all communications K (1) k : the sum of the γ i that cross D (1) k In this case, K (1) = K for each k k P XY = ( p + q ) × K α Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 11 / 34

  4. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p ) . K : sum of all communications K (1) k : the sum of the γ i that cross D (1) k In this case, K (1) = K for each k k P XY = ( p + q ) × K α Lower bound on P max . Ideal sharing of one communication: Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 11 / 34

  5. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p ) . p − 1 � K (1) � α q − 1 � K (1) � α � � k k P max ≥ 2 k + (2 p − 1) 2 k 2 p − 1 k =1 k = p q + p − 2 � � α K (1) � k + 2( q + p − k − 1) , 2( q + p − k − 1) k = q Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 11 / 34

  6. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p ) . p − 1 � K (1) � α q − 1 � K (1) � α � � k k P max ≥ 2 k + (2 p − 1) 2 k 2 p − 1 k =1 k = p q + p − 2 � � α K (1) � k + 2( q + p − k − 1) , 2( q + p − k − 1) k = q � p k =1 k 1 − α ≥ K (1) = K and � p − 1 1 dx / x α − 1 k Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 11 / 34

  7. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p ) . p − 1 � K (1) � α q − 1 � K (1) � α � � k k P max ≥ 2 k + (2 p − 1) 2 k 2 p − 1 k =1 k = p q + p − 2 � � α K (1) � k + 2( q + p − k − 1) , 2( q + p − k − 1) k = q � p k =1 k 1 − α ≥ K (1) = K and � p − 1 1 dx / x α − 1 , hence k � 1 1 q − p � � 1 − p 2 − α � P max ≥ K α 2 × + . 2 α − 1 2 − α (2 p − 1) α − 1 Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 11 / 34

  8. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p ) . p − 1 � K (1) � α q − 1 � K (1) � α � � k k P max ≥ 2 k + (2 p − 1) 2 k 2 p − 1 k =1 k = p q + p − 2 � � α K (1) � k + 2( q + p − k − 1) , 2( q + p − k − 1) k = q � p k =1 k 1 − α ≥ K (1) = K and � p − 1 1 dx / x α − 1 , hence k � 1 1 q − p � � 1 − p 2 − α � P max ≥ K α 2 × + . 2 α − 1 2 − α (2 p − 1) α − 1 Altogether, P max = O ( K α ) and P XY = O ( p × K α ), hence the result. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 11 / 34

  9. Framework Theoretical results Heuristics Simulations Manhattan vs XY; single source and destination Theorem The upper bound of P XY / P max in O ( p ) is tight. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 12 / 34

  10. Framework Theoretical results Heuristics Simulations Manhattan vs XY; multiple sources and destinations Theorem Given that q = O ( p ) , an upper bound of P XY / P max is in O ( p α − 1 ) . Theorem The upper bound of P XY / P max in O ( p α − 1 ) can be achieved with a 1 - MP routing on a square CMP. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 13 / 34

  11. Framework Theoretical results Heuristics Simulations NP-completeness of Manhattan routing Theorem Finding a s- MP routing that minimizes the total power consumption while ensuring that link bandwidths are not exceeded is a NP-complete problem. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 14 / 34

  12. Framework Theoretical results Heuristics Simulations Outline of the talk Framework 1 Theoretical results 2 Heuristics 3 Simulations 4 Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 15 / 34

  13. Framework Theoretical results Heuristics Simulations Summary of the heuristics Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 16 / 34

  14. Framework Theoretical results Heuristics Simulations Summary of the heuristics Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 16 / 34

  15. Framework Theoretical results Heuristics Simulations Summary of the heuristics Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Two-bend ( TB ): for each communication, chooses the best path with two bends. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 16 / 34

  16. Framework Theoretical results Heuristics Simulations Summary of the heuristics Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Two-bend ( TB ): for each communication, chooses the best path with two bends. XY improver ( XYI ): starts from XY assignment, and moves communications from the highest loaded link. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 16 / 34

  17. Framework Theoretical results Heuristics Simulations Summary of the heuristics Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Two-bend ( TB ): for each communication, chooses the best path with two bends. XY improver ( XYI ): starts from XY assignment, and moves communications from the highest loaded link. Path remover ( PR ): virtually pre-assigns communications onto links, and iteratively prevents communications from using highly loaded links. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 16 / 34

  18. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Two-bend ( TB ): for each communication, chooses the best path with two bends. XY improver ( XYI ): starts from XY assignment, and moves communications from the highest loaded link. Path remover ( PR ): virtually pre-assigns communications onto links, and iteratively prevents communications from using highly loaded links. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 18 / 34

  19. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  20. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  21. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  22. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  23. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  24. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  25. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  26. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  27. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  28. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  29. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  30. Framework Theoretical results Heuristics Simulations Simple greedy ( SG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 19 / 34

  31. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Two-bend ( TB ): for each communication, chooses the best path with two bends. XY improver ( XYI ): starts from XY assignment, and moves communications from the highest loaded link. Path remover ( PR ): virtually pre-assigns communications onto links, and iteratively prevents communications from using highly loaded links. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 20 / 34

  32. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  33. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  34. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  35. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  36. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  37. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  38. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  39. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  40. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  41. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  42. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  43. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  44. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  45. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  46. Framework Theoretical results Heuristics Simulations Improved greedy ( IG ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 21 / 34

  47. Framework Theoretical results Heuristics Simulations Two-bend ( TB ) Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Two-bend ( TB ): for each communication, chooses the best path with two bends. XY improver ( XYI ): starts from XY assignment, and moves communications from the highest loaded link. Path remover ( PR ): virtually pre-assigns communications onto links, and iteratively prevents communications from using highly loaded links. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 22 / 34

  48. Framework Theoretical results Heuristics Simulations Two-bend ( TB ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 23 / 34

  49. Framework Theoretical results Heuristics Simulations Two-bend ( TB ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 23 / 34

  50. Framework Theoretical results Heuristics Simulations Two-bend ( TB ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 23 / 34

  51. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Two-bend ( TB ): for each communication, chooses the best path with two bends. XY improver ( XYI ): starts from XY assignment, and moves communications from the highest loaded link. Path remover ( PR ): virtually pre-assigns communications onto links, and iteratively prevents communications from using highly loaded links. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 24 / 34

  52. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  53. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  54. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  55. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  56. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  57. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  58. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  59. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  60. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  61. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  62. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  63. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  64. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  65. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  66. Framework Theoretical results Heuristics Simulations XY improver ( XYI ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 25 / 34

  67. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Simple greedy ( SG ): greedily assigns communications, hop by hop, on the least loaded link. Improved greedy ( IG ): virtually pre-assigns communications onto links, then almost like SG . Two-bend ( TB ): for each communication, chooses the best path with two bends. XY improver ( XYI ): starts from XY assignment, and moves communications from the highest loaded link. Path remover ( PR ): virtually pre-assigns communications onto links, and iteratively prevents communications from using highly loaded links. Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 26 / 34

  68. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 27 / 34

  69. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 27 / 34

  70. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 27 / 34

  71. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 27 / 34

  72. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 27 / 34

  73. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 27 / 34

  74. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 27 / 34

  75. Framework Theoretical results Heuristics Simulations Path remover ( PR ) Pitt Benoit, Melhem, Renaud, Robert Power-aware Manhattan routing on CMPs 27 / 34

Recommend


More recommend